|
Synthesia. Yah, a new unicorn! AI-based video creation platform Synthesia has raised $90M at a $1B valuation from investors led by VC firm Accel and backed by U.S. chipmaker Nvidia and others.
2020 Visual 1st presenter Synthesia empowers users to make their own video avatars, which look and sound like humans and speak the user-provided text in more than 120 languages. Use cases include corporate presentations, training videos, and personalized video greetings.
Printbox. Generative AI. Tomorrow Visual 1st sponsor Printbox is slated to unveil Masterpiece AI, a generative AI solution for printing, retail, and e-commerce industries.
PiktID. Swapping real faces with synthetic ones. AI-based PiktID lets you replace faces in your photos with synthetically created faces. If that’s not enough, it also lets you modify faces, e.g. by letting them smile. A deepfake toy for kids? Not PiktID’s target – brands can save on model releases by anonymizing photos, localizing a single photo for different cultures or campaign targets, and feature children without privacy concerns.
Meta. Good for a headline: “Human-like” AI image creation. Meta is planning to provide researchers with access to components of a new AI model, I-JEPA, which uses background knowledge about the world to fill in missing pieces of images. Meta’s model is said to be able to analyze and complete unfinished images more accurately than existing models that fill in the gaps by looking at nearby pixels. No more AI hands with 6 fingers!
Google Shopping. Virtual try-ons. Google Shopping is launching a new generative AI shopping feature that visualizes clothes on a lineup of real-life fashion models. Google’s virtual try-on tool for apparel takes an image of clothing and attempts to predict how it would drape, fold, cling, stretch and form wrinkles and shadows on a set of models in different poses. For now, U.S. shoppers can virtually try on women’s tops from brands including Anthropologie, Everlane, H&M and LOFT. Alas, men’s tops will launch later in the year.
Mylio. Upping its AI photo management features. Mylio announces a range of AI-powered features for its photo management software Mylio Photos. These include QuickFilters to filter photos for specific metadata, Photo DeClutter to identify the best photos from photos captured as bursts, and Social Media Recovery to back up photos and videos previously shared on social media sites. Mylio Photos’ AI tools all run on the user’s local devices, not in the cloud.
vAIsual. Elephants, Birds and Dogs! Oh my! Last year’s Visual 1st presenter vAIsual announces 3 non-biometric datasets, consisting of over 130K images of elephants, dogs, and birds. The largest one, the Diverse Dogs Collection with over 110K images, can be used for training AI models to detect dogs in images, recognize breeds, analyze dog behaviors, and contribute to various applications in veterinary science, pet adoption, and animal welfare, according to vAIsual. Its datasets include detailed and consistent metatags, come in various resolutions, and, most importantly, have legal clearances.
Vimeo. Adding AI video editing tools. Vimeo is adding a suite of AI tools aimed at making video creation and editing easier. They include an AI script generator, a teleprompter so you can read the script while recording your video, and a text-based video editor, which allows you to cut filler words, long pauses, and other unwanted bits automatically with a single click of a button.
Canva. It’s time to extend that platform. What’s next when you’ve become a unicorn and have amassed 135M monthly active users? You court developers to extend your platform by addressing use cases that are too specialized or niche to address yourself. Canva launched it is Canva Developers Platform, offering its Canva Apps SDK, which grants access to resources such as javascript libraries, documentation, sample apps, UI guides and, notably, the new Canva Apps APIs. And to get things rolling, Canva also established a $50M Canva Developers Innovation Fund.
Stability AI. Uncropping. Uncrop is not in my dictionary, but it perfectly indicates the reverse of cropping: you enlarge the image by letting Stable Diffusion’s AI fill in the blanks of what surrounded the original image. (Uncrop is reminiscent of DALL-E’s Outpainting and Photoshop’s Generative Fill features). You’ve taken a photo that was a bit too zoomed in? Or changed your mind and want the image with a different aspect ratio? Uncropping is the answer.
Google Photos. Face recognition when there are no visible faces. Google has taken Photos' face recognition prowess to the next level, allowing the app to recognize people's faces in photos even when they are looking completely away from the camera, such as when only the back of someone’s head is visible. How does Google Photos do this? According to the company, its machine learning models have been improved to “group people based on clothing and other visual cues across photos taken within a similar timeframe.”
Given Google Photos’ top-flight engineering team and access to a trove of cutting-edge company-owned technologies, how does Google Photos decide which features to develop, and which ones to pass on? Hear from John Fisher, Sr. Director of Engineering at Google Photos, in a special fireside chat session at Visual 1st.
| 
